F-VLM: open-vocabulary object detection upon frozen vision and language modelsai.googleblog.com3 pointsshantanu_sharma3 years ago